Hyperspectral Anomaly Detection with CNN-Based VAE and RX Algorithm

Hyperspectral imaging (HSI) offers rich spectral and spatial information, making anomaly detection in hyperspectral images a critical and widely applied task. While combining Variational Autoencoders (VAE) with the Reed-Xiaoli (RX) algorithm has shown advantages, existing methods often rely on individual pixels without fully exploiting spatial information. In this work, we propose a CNN-based VAE framework that extracts latent representations from multi-scale data cubes, which are then analyzed by the RX algorithm to detect anomalies. This method effectively incorporates both spectral and spatial information, addressing the limitations of existing detectors. Experiments on multiple datasets demonstrate the superior accuracy and efficiency of our approach in detecting anomalies in hyperspectral images